Large Language Models (LLMs) have recently been leveraged for asset pricing tasks and stock trading applications, enabling AI agents to generate investment decisions from unstructured financial data. However, most evaluations of LLM timing-based investing strategies are conducted on narrow timeframes and limited stock universes, overstating effectiveness due to survivorship and data-snooping biases. We critically assess their generalizability and robustness by proposing FINSABER, a backtesting framework evaluating timing-based strategies across longer periods and a larger universe of symbols. Systematic backtests over two decades and 100+ symbols reveal that previously reported LLM advantages deteriorate significantly under broader cross-section and over a longer-term evaluation. Our market regime analysis further demonstrates that LLM strategies are overly conservative in bull markets, underperforming passive benchmarks, and overly aggressive in bear markets, incurring heavy losses. These findings highlight the need to develop LLM strategies that are able to prioritise trend detection and regime-aware risk controls over mere scaling of framework complexity.

Multi-Agent Reinforcement Learning (MARL) has shown promise in solving complex problems involving cooperation and competition among agents, such as an Unmanned Surface Vehicle (USV) swarm used in search and rescue, surveillance, and vessel protection. However, aligning system behavior with user preferences is challenging due to the difficulty of encoding expert intuition into reward functions. To address the issue, we propose a Reinforcement Learning with Human Feedback (RLHF) approach for MARL that resolves credit-assignment challenges through an Agent-Level Feedback system categorizing feedback into intra-agent, inter-agent, and intra-team types. To overcome the challenges of direct human feedback, we employ a Large Language Model (LLM) evaluator to validate our approach using feedback scenarios such as region constraints, collision avoidance, and task allocation. Our method effectively refines USV swarm policies, addressing key challenges in multi-agent systems while maintaining fairness and performance consistency.

In this article, a new input-referred second-order intercept point (IIP2)-calibration-free receiver (RX) front-end employing mixer-sharing N-path notch filter feedback is proposed for 5G new radio (NR) cellular applications. An RX front-end architecture with a high-Q N-path RF filtering low-noise amplifier (LNA) enhances the overall IIP2 performance of the RX via the blocker filtering. However, the local oscillator (LO) chain burden increases with the number of switches in the N-path filter. The main down-conversion mixers are reused as part of switches for the N-path notch filter, reducing the LO chain burden and current consumption. The RX front-end comprises a balun-LNA, main down-conversion passive mixers driven by 25% duty-cycle LO signals, isolation resistors, and transimpedance amplifiers. Additional high-pass filters, up-conversion mixers, and reused down-conversion mixers form the global N-path notch filter feedback. Fabricated using a 65 nm CMOS process, the RX front-end was primarily characterized for the mid band of 5G NR sub-6 GHz cellular applications. It achieved a noise figure of 3.67 dB, a conversion gain of 44.7 dB, and an out-of-band input-referred third-order intercept point of 15.35 dBm. Additionally, an IIP2 of >72 dBm was obtained without calibration. The RX front-end drew a 13.83 mA bias current from a 1 V nominal supply voltage, with a 0.45 mm2 active die area.

Most of the wearable hand kinesthetic (force) feedback systems have only been evaluated from a physical perspective, i.e., motion measurement, force transmission, system stability, etc. However, by taking this approach, we cannot gather sufficient data to evaluate the performance of wearable kinesthetic feedback systems as haptic devices. In this paper, we evaluated the wearable hand kinesthetic feedback system developed in our laboratory from psychophysical and user experience (UX) perspectives. We designed the experiments considering the characteristics of our wearable hand system. For psychophysical evaluation, ten subjects participated in force detection, discrimination, and identification experiments. The force detection threshold of our system was better than those of the commercialized haptic systems, and discrimination threshold was similar with the human perception threshold. In the identification experiment, the subjects could distinguish the size and shape of virtual objects over 2.5 levels. For UX evaluation, 15 subjects answered a questionnaire after conducting a task using our wearable hand system in laboratory environment. The UX results verified that the subjects were satisfied with our system by giving a score of 5.87 out of 7 points. It also concluded that our system can provide a good VR experience to users by achieving over 5.5 points in ease of use, learnability, and utilitarian value.